🎯 Prompt injection attack

The items that name it 3

• LTBD: Learnable Trust-Boundary Delimiters for Prompt Injection Defense
  paper2026-10-08It is written here as Prompt injection attack: malicious instructions embedded in external data override user intent.
• RAG-PIBench: A Leakage-Aware Benchmark for Prompt-Injection Detection in Trustworthy RAG Systems
  paper2026-10-06It is written here as Prompt-injection attack.

  Attacks embedded in retrieved content that exploit RAG systems

• PISmith: Reinforcement Learning-based Red Teaming for Prompt Injection Defenses
  paper

  Attacker embeds malicious instructions into an LLM's input context (e.g., web content or documents) to manipulate the LLM's output; core threat addressed by the paper